Network Device Performance Prediction Using Historical Data

The system performs preliminary actions by collecting historical performance data and training the prediction model in advance. The model is trained offline using historical data from multiple network devices, and the trained model parameters are stored for later use in generating predictions. This preliminary training phase enables accurate real-time predictions without requiring complex computations during the actual prediction moment.

The patent introduces a prediction model as an intermediary between the manual estimation approach and the actual performance measurement. This model acts as a mediator that processes input parameters (device specifications, configuration settings, traffic patterns) and generates predicted performance consumption values. The model bridges the gap between simple input data and accurate performance predictions, resolving the contradiction between simplicity and accuracy.

The prediction system is segmented into distinct functional modules: a data collection module that gathers historical performance data, a model training module that processes the data and generates prediction models, and a prediction generation module that uses the trained models to produce predictions. This segmentation allows each module to be optimized independently and simplifies the overall system architecture while maintaining high prediction reliability.

The system utilizes parameter changes in historical performance data to train the prediction model. By analyzing variations in device configurations, traffic patterns, and performance metrics over time, the model learns to predict future performance consumption accurately. The model adapts to different parameter combinations and their impact on performance, enabling reliable predictions across diverse network scenarios.

The system extracts only the most relevant features and parameters from the comprehensive device information and historical data. Instead of processing all available data, the prediction model identifies and extracts key features such as device type, configuration settings, traffic patterns, and performance metrics that have the greatest impact on performance consumption. This extraction process reduces data processing complexity while maintaining prediction precision.

The system performs preliminary data processing and feature extraction during the model training phase. Historical data is pre-processed, cleaned, and transformed into suitable formats before being used to train the prediction model. This preliminary action reduces the complexity of real-time data processing during prediction operations, as the model has already learned from the pre-processed historical data.